- Email: [email protected]
Rental Adjustment and Valuation in Overbuilt Markets: Evidence from the Sydney Office Market
JOURNAL OF URBAN ECONOMICS ARTICLE NO.
39, 51]67 Ž1996.
0003
Rental Adjustment and Valuation in Overbuilt Markets: Evidence from the Sydney Office MarketU PATRIC H. HENDERSHOTT Max M. Fisher College of Business, The Ohio State Uni¨ ersity and NBER, 1775 College Road, Columbus, Ohio 43210 Received May 9, 1994; revised August 29, 1994
In equilibrium, real estate will be valued at Žapproximately. replacement cost. When value exceeds replacement cost in a market, new construction surges, raising vacancies and lowering rents and thus value. When replacement cost exceeds value, replacement construction is deferred, lowering vacancies and raising rents and value. Moreover, expectations of the construction, vacancy, and rent responses imply that existing real estate values will be only modestly sensitive to changes in required real after-tax returns caused by real interest rate or tax changes.1 Because higher long-term real interest rates lower real estate values, construction declines and rental income on existing properties rises. That is, the rental cash flows on existing real estate are more analogous to those on a sluggishly adjusting adjustable-rate loan than to the fixed coupons on a long-term bond; thus the percentage real estate value decline would be far less than that of a long-term bond. U
I thank BOMA for supplying me with data, Bond University for making facilities available to me while the first draft of the paper was being written, and Frank Russell Australia for financial support. The cooperation of the Sydney JLW Research and Consultancy Office, which supplied me with real rental data Žincluding rent concessions. and vacancy rate projections of market participants, was crucial to the success of this project. Individuals supplying useful input include Adrian Harrington of BOMA, David Dickinson, N. H. Seek, and S. G. Slim of the Sydney JLW Research and Consultancy Office, John Bowers, Ralph Jackson, and David MacKenzie of Frank Russell, and Barry Reece of the University of NSW. I also thank two anonymous reviewers and the journal editor. Earlier versions of this paper were presented at the Univ of ConnrAREUEArFannie Mae 1993 International Real Estate Conference, Mystic, CT, October 1993 and the first Global Finance Conference in Monterey, CA, April 1994. 1 Hendershott, Follain, and Ling w3x make this point in their analysis of the impact of the 1986 tax act on real estate values. Ling w8x extends the argument in his analysis of the impact of capital gains tax rate changes. 51 0094-1190r96 $12.00 Copyright Q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.
52
PATRIC H. HENDERSHOTT
As a result of the construction response mechanism, deviations of real estate values greater than five percent from replacement cost would be rare in relatively stable markets, i.e., those without substantial excess supply or demand. Unfortunately, stable well-functioning is not the way one would describe office markets in many parts of the western world. In London, Tokyo, most national capitals of Scandinavia, most state capitals of Australia, and many large U.S. cities, vacancy rates of 20 to 30% have been common, real effective rents have plunged, and values are almost certainly far below replacement cost. In the transforming socialist economies of Central and Eastern Europe, the profit-motivated construction equilibrating mechanism described above has not existed for at least a half a century. For the value of any type real estate in any locale there to be near replacement cost would be pure happenstance. This paper illustrates a method for valuing real estate in overbuilt Žor underbuilt. markets with office building data from Sydney, Australia. The method is presented in Section I, where fundamental value is expressed as replacement cost less the present value of expected below-equilibrium real rental income. Underpinnings for the calculation of below-equilibrium real rents are given in Section II; equilibrium real rents are specified and real rental adjustment equations are estimated using data from the 1970]92 period. The percentage change in real effective rents is related to gaps between both the natural and actual vacancy rates and equilibrium and actual real rental rates. Expectations of future equilibrium real rents Žincluding real discount rates. and vacancy rates thus provide a future rental income stream from which the present value of expected belowequilibrium rental income, and thus fundamental value, can be computed. The valuerreplacement-cost ratio is calculated for Sydney in Section III, and its sensitivity to a number of assumptions is indicated. Most importantly, the 1992 ratio is not sensitive to real interest rate Žand thus risk premium. assumptions due to offsetting adjustments in below equilibrium rents and the discount rate. Also, the value of above-market leases, which likely existed on Sydney buildings leased up in the early 1990s, is computed. In the absence of price ‘‘bubbles,’’ price should equal fundamental value plus the value of above-market leases. 2 In Section IV, estimates of the ratio of fundamental values plus above-market leases to replacement cost are compared and contrasted with ratios derived from series commonly reported in two prominent Australian real estate sources, JLW and
See Flood and Hodrick w2x, Shleifer and Summers w10x and Stiglitz w11x for useful discussions of the literature on asset market bubbles. While the construction response will not guarantee the absence of bubbles, it will clearly limit their duration. 2
53
REAL ESTATE IN OVERBUILT MARKETS
BOMA.3 During the 1990]92 period, the JLW and BOMA series were 50 to 100% greater than fundamental value Žplus the value of enforceable above-market leases.. An important issue is whether the reported series are misrepresentations of market value or reflect a speculative bubble. While independent evidence on Sydney land values is offered for the development of a speculative bubble in the second half of the 1980s, the same evidence implies the bubble burst in the early 1990s. Further, the surge in Sydney office market construction in the late 1980s, a reasonable builder response to a prices exceeding fundamental value, almost guarantees the bursting of such a bubble in the early 1990s. That is, the 1992 BOMA data Žand the JLW data to a lesser extent. appear to overstate market value significantly. 4 I. THE BASIC VALUATION MODEL The fundamental value of a property in period t Ž Vt . is the present value of expected future net operating income. Assuming, for simplicity, that the expected ratio of operating expenses to replacement cost Žoper. and one-period real risky financing rates Ž r . are constant through time, value at time t for a tax-exempt investor is
Vt s
`
Ý js1
g tqj Ž 1 y ¨ tqj . y exp RCt Ž 1 y d .
Ž 1 q rtU .
j
j
,
Ž 1.
where g is the expected real gross rental rate per replacement-cost dollar of occupied space, ¨ is the expected vacancy rate, and RC, the current replacement cost of built space, is expected to decline at the depreciation rate, d.5 3 The Sydney Jones]Lang]Wootten ŽJLW. Research and Consultancy Office reports effective rent, vacancy rate, and capital value series for most Australian capital cities. The Building and Operating Management Association ŽBOMA. reports cash flow and appreciation returns for different property types in most capital series. The BOMA series are analogous to those produced by the Frank Russell company in the United States Žthe appreciation return component is based on appraisals, not transactions .. 4 Some have argued that appraisers are reluctant to revalue properties sharply over short periods. This tendency could be exaggerated for downward adjustments due to the incentives of investment managers to maintain the values upon which their percentage fees are based. 5 Technically, the zero-coupon rate Žplus an appropriate risk premium. for each maturity should be used to discount each period’s cash flow. When the yield curve is steeply sloped and the cash flows are markedly uneven, using different-period discount rates can greatly affect valuation. Similarly, the expected inflation rate between the current period and when the cash flow will be received should be used to compute each cash flow.
54
PATRIC H. HENDERSHOTT
In equilibrium, ¨ tqj will equal ¨ U for all j, gross rents will equal real financing costs, economic depreciation, and operating expenses, and value will equal replacement cost. More specifically, if g Ž 1 y ¨ U . s r U q d q exp s g U
Ž 2.
and ¨ s ¨ U Žthe natural vacancy rate., then value will equal replacement cost. The validity of Ž1. can be verified by setting all ¨ tqj s ¨ U , all g tqj s g Ut , and substituting Ž2. into Ž1.. The determinants of the VrRC ratio are of particular interest to us. To place these determinants in better focus, we add and subtract `
Ý
Ž rt q d . RCt Ž 1 y d . Ž 1 q rt .
js1
jy1
s RCt
j
from the right hand side of Eq. Ž1.. Combining terms and dividing by RC, we see that the VrRC ratio is reduced from unity to the extent that expected future below-equilibrium rents ŽBERI. exist: Vt RCt
s1y
BERI t RCt
,
Ž 3.
where BERI t RCt
N
s
Ý js1
j Ž rtU q d q exp. y g tqj Ž 1 y ¨ tqj . Ž 1 y d .
Ž 1 q rtU .
j
.
Ž 4.
The model has been set up in terms of tax-exempt investors. A valuation model for taxable investors can be obtained by introducing taxes into Eqs. Ž1. and Ž2..6 Regarding Eq. Ž1., the cash flows would be taxed, an after-tax discount rate would be used, and a term to reflect the value of tax depreciation would be added. In addition, equilibrium net rents in Eq. Ž2. would depend upon individual tax rates and tax depreciation. However, in an overall equilibrium, net rents for all investor classes will be equal; the required after-tax returns and portfolio shares invested in real estate will differ across investor classes, with more heavily taxed investors having smaller portfolio shares and thus smaller risk premiums.7 In disequilibrium, properties will shift toward whatever investors have the larger valuation. The required returns of these investors will rise and returns for See Hendershott and Kane w5x for full presentation of such a model. For analysis of a portfolio model in which investors in different tax brackets and with different risk-aversion parameters}and thus different required returns}hold different portfolio shares, see Hendershott and Won w6x. 6 7
REAL ESTATE IN OVERBUILT MARKETS
55
the selling investors will fall until all investors stand again in equilibrium, with properties valued equally by all investor classes. That is, appropriately specified and parameterized, taxable and tax-exempt models will yield the same valuations. The keys to valuation are first specifying equilibrium rents Žthe real interest and depreciation rates and the operating expense ratio in the tax-exempt model. and second determining both how far the rental rate Ž g adjusted for vacancies. is currently below the equilibrium rate and the time path along which the rental rate will return to equilibrium. The latter involves projecting the time path of vacancy rates and the real gross rental share. While we have estimates of market participants’ expectations of future vacancy rates, expected real gross rents are based upon an estimated real gross rental adjustment equation. This is the topic of the next section. II. EXPLANATION OF SYDNEY REAL EFFECTIVE RENTS A long tradition exists of relating the percentage change in real effective rents to the difference between the ‘‘natural’’ and observed vacancy rates w1, 9, 14x: D g tqjrg tqjy1 s lŽ ¨ U y ¨ tqjy1 . .
Ž 5.
As popular as this reduced-form relation has been, it suffers from three problems. First, the equation provides no relationship between the le¨ els of actual and equilibrium rents, which must be equal in equilibrium. Second, the adjustment equation requires substantial overshooting of the natural rate in response to shocks other than a change in the equilibrium rental rate.8 Third, the equation cannot hold simultaneously for leases of different terms. To illustrate these problems, consider a market initially in equilibrium with gross rents equal to equilibrium rents, the vacancy rate equal to the natural rate, and value equal to replacement cost. A negative demand or positive supply shock Žexcess space unexpectedly becomes available. increases vacancies and decreases rents Žand thus values.. The contraction in construction resulting from value declines turns vacancies around, but rents continue to fall until the vacancy rate is back to the natural rate. The market must then go through a reverse adjustment with rents rising, first with falling vacancies and later with rising vacancies, until they return to equilibrium, vacancies return to the natural rate, and value again equals Wheaton and Torto w13x discuss the microfoundations of the adjustment equation and estimate a less constrained form for numerous U.S. cities on data from the 1979]91 period. They do not impose an equilibrium rent relationship. 8
56
PATRIC H. HENDERSHOTT
replacement cost. To create the force returning rents to equilibrium, the rental adjustment equation necessarily entails the vacancy rate substantially overshooting the natural rate. The vacancy rate will decline below the natural rate because value is less than replacement cost Žnet construction is negative. when the actual vacancy rate first returns to the natural rate. Moreover, even if the rental adjustment equation characterized oneperiod leases, it cannot hold for longer term leases. With rational expectations, the rent on multiperiod leases will be an average of the expected future rents on one-period leases, just as longer term bond rates are averages of future one-period bond rates. Thus expectations that oneperiod rents will rise in the future will turn rents on multiperiod leases upward before the vacancy rate reaches its natural level.9 Logically, Eq. Ž5. cannot hold simultaneously for leases of different term. To allow a more general adjustment path with pleasing long-run properties, real effective rents are specified as adjusting to gaps between both the natural and actual vacancy rates Ž ¨ U and ¨ y1 . and equilibrium and actual gross rents Ž g U and gy1 .: D g tqjrg tqjy1 s lŽ ¨ U y ¨ tqjy1 . q b Ž g Utqj y g tqjy1 . .
Ž 6.
With this equation, vacancy rates do not have to overshoot following a supply shock. After high vacancy rates have dragged rents significantly below equilibrium, the known eventual return to equilibrium acts as a force causing real rents to rise, even when the vacancy rate is still above the natural rate. Moreover, with ¨ s ¨ U , real rents will adjust to their equilibrium value. The Data Rental adjustment equations are estimated using annual data from Sydney over the 1970]92 period.10 The natural vacancy rate is treated as a constant over time.11 In Ž5. or Ž6., ¨ tqjy1 is included as a regressor and the constant term is l¨ U . Thus the natural rate can be computed after the 9
In a cross-sectional study of individual leases, the term of the lease can be entered as a regressor to capture the differential impact of future expected values of one-term leases on the value of longer-term leases with different terms. In a pooled time-series cross-sectional study, the impact of term should be allowed to vary over time because the term structure of leases varies over time. 10 All data were generously supplied by the Sydney JLW Research and Consultancy office w7x. 11 Few degrees of freedom prevent serious testing of a nonconstant natural vacancy rate. It is worth noting, though, that reported evidence seeming to suggest large time variation in natural vacancy rates may simply reflect continuing market disturbances ŽHendershott and Kane, w5x..
REAL ESTATE IN OVERBUILT MARKETS
57
estimation as the ratio of the constant term to the coefficient on the lagged vacancy rate Ž l.. Real effective rents require the downward adjustment of real face rents for both free rent periods and tenant improvements. Conveniently for us, JLW w7x has collected data on both of these terms for Sydney office markets over the 1970]91 period. According to their data, rent incentives increased sharply over the 1989]91 period, rising from less than 4 months equivalent free months rent Žincluding adjustment of improvements. on a ten-year lease to almost 23 free months. JLW also compute effective real rents by discountingramortizing real cash flows with a real interest rate. We have adjusted JLW’s methodology in only one respect; they amortize the present value of the rent incentives over the life of the building whereas the correct amortization period is the length of the lease. With this adjustment, the amortized value of the rent incentives jumped from 6% of real face rent in 1989 to 41% in 1992.12 The percentage change in the resultant real effective rent is the dependent variable in our analysis. The equilibrium real gross rental rate equals the real risk-free rate plus a risk premium, the depreciation rate, and the expense ratio. An approximate Australian risk-free real rate was computed as Ž1 q r f .rŽ1 q p . y 1 for semi-annual intervals over the December 1969 to June 1992 period, using the 10-year Treasury rate as the risk-free rate Ž r f . and a three-period average of annualized percentage changes in the Deflator for Private Final Consumption Expenditures as the expected inflation proxy Žp .. The series actually employed is a smoothed version of the calculated series because longer term real interest rates should move relatively smoothly. The smoothed real risk-free rate drops from 0.02 in 1970 to zero in early 1971, stays there until the middle of 1974, abruptly plunges to y0.04, and remains there through the middle of 1977. The series then rises linearly to 0.06 in June 1982, where it stays through June 1992. The risk premium and depreciation rates are set at 0.035 and 0.025, respectively, throughout the period. BOMA data indicate that the ratio of outgoings Žoperating expenses. to gross rates has ranged from 0.285 to 0.335; oper is set equal to 0.05, giving a gross rental rate of 0.17 and an outgoings ratio of 0.294 Ž0.05r0.17.. Thus the equilibrium rental rate declines from 0.13 in 1970 to 0.07 in the 1975]77 period and rises to 0.17 throughout the 1982]92 span. To convert this gross rental rate series to a current Ž1992. dollar rent per square meter, the series is multiplied by the real rent level in a year in which equilibrium and actual rents appear to have been equal. Given the steadiness of both actual and equilibrium rents 12
Their methodology understates the increase in value of rent incentives and thus the decrease in effective rents between 1989 and 1992. Rather than a 48% real decline, they compute a 26% decline.
58
PATRIC H. HENDERSHOTT
during the 1983]85 period, real effective rents are assumed to have equaled equilibrium rents in June 1986. This equilibrium rent series and the JLW real effective rent series are plotted in Fig. 1. A clear relationship is evident, with the JLW series generally lagging the equilibrium series. The last key input is the vacancy rate. This series began and ended the 1970s at about 3% with a sharp increase and reversal in between Žthe peak was 13% at the end of 1976.. Real effective rents followed the vacancy rate with about a 2 year lag. The vacancy rate stayed below 5% throughout the 1980s and was 3% in the middle of 1989; during the second half of the decade, real effective rents rose by 50%. Since then a plethora of completions and negative absorption has driven vacancy to 23% and halved real effective rents. The Estimates Table 1 reports some estimates of the rental adjustment equation. Coefficient estimates and standard errors Žin parentheses ., the equation R 2 Žnot adjusted for degrees of freedom., and the implied natural vacancy rate are reported. The first estimate is of the naive model in which rents adjust solely to the difference between the natural and actual vacancy rates. As is seen, the equation explains less than a third of the variance. The implied natural vacancy rate is 6.0%. Adding the equilibrium rent gap more than doubles the explanatory power. While the explanatory power is reasonable, the equation misses a significant portion of the broad swings in real effective rents. More specifically, between 1978 and 1989 ŽJune. real effective rents per square meter rose by $513, $375 of which was lost by June 1992. The equation explains only 72 and 38%, respectively, of the increase and reversal.
FIG. 1. Actual and equilibrium rents.
59
REAL ESTATE IN OVERBUILT MARKETS
TABLE 1 Explanation of the Percentage Change in Real Effective Sydney Office Market Rents, 1971]92 Equ.
Constant
¨ y1
1.1
.1125
1.2
.1121
1.3
.1159
y.0189 Ž.0062. y.0176 Ž.0047. y.0171 Ž.0046.
g U y gy1
2.57 Ž0.60. 2.26 Ž0.64.
D¨
y.0085 Ž.0069.
R2
¨U
.317
6.0
.650
6.4
.679
6.8
Note. The data are from JLW. Vacancy rates are in percentage points; rental rates are in decimals. Standard errors are in parentheses. See the text for exact definition of the variables.
To better explain the sharp fall in real rents during the June 1989]June 1992 period, the forward change in the vacancy rate Žthe December to December change explaining the June to June change in rents. has been added. Knowing that substantial capacity was coming on line should have directly reduced real effective rents. While adding the forward vacancy-rate change does not increase the explanatory power much Žthe t-ratio on the variable is barely above one., the equation is far better able to explain the pre-1989 rise in real rents and especially the subsequent decline. Rather than the 72 and 38% just mentioned, 85% of the increase and 63% of the decrease are explained. The estimate of the natural vacancy rate rises slightly to 6.8%. Figure 2 plots the level of real effective rents and the level predicted dynamically Žthe initial level moved forward based on the predicted changes.. Because the annual errors in predicted changes are largely independent, the predicted level tracks the actual level reasonably well. Figure 2 also includes forecasted rent levels for the 1993]2003 period, assuming the Sydney vacancy rate is 23% through December 1994 and then declines Žsee below.. The equation predicts another $120 Ž27%. decline in real effective rents before an upturn in 1998.13 III. VALUE r REPLACEMENT-COST RATIOS IN JUNE 1992 In this section, the model is fully parameterized and VrRC ratios are computed for Sydney in June 1992 under the base assumptions using each of the real rental adjustment equations. The sensitivity of the calculated 13 According to recent JLW data, real effective rents declined by 30% between June 1992 and June 1993.
60
PATRIC H. HENDERSHOTT
FIG. 2. Actual and predicted real effective rents.
VrRC ratio to each of a more optimistic vacancy rate decline, alterations in estimated rental-adjustment coefficients, and a decrease in real interest rates is illustrated. Last, the value of long-lived enforceable leases, which would result in slower declines in real rents after 1989, is also computed. Model Parameterization The VrRC ratio is unity less the present value of below-equilibrium future cash flows divided by replacement cost. As can be seen from Eqs. Ž2. and Ž4., the variables to be specified are: the real risky discount rate Ž r ., the depreciation rate Ž d ., the operating expense ratio Žoper., the natural vacancy rate Ž ¨ U ., and the time paths of future vacancy and gross rental rates. Moreover, from Eq. Ž6., the time path of future gross rental rates is fully specified by the other variables. With constant future values of r, d, ¨ U , and oper, only a future time path of vacancy rates is needed for the calculations. Most of the June 1992 values have already been specified: r s 0.095, d s 0.025, and oper s 0.05, giving a gross equilibrium rental rate of 0.17. From the estimation, ¨ U s 0.068. With the assumption of equality between actual and equilibrium rents in June 1986, the actual gross rental rate in June 1992 is 0.117. The last key assumption is the future time path of vacancy rates expected by market participants. Table 2 lists assumed future vacancy rates in Sydney as of year-end over the 1985]92 period. Each column contains the rate expected during June of the year heading the column to exist at the end of the years in the far left column Žthe diagonal contains observed
61
REAL ESTATE IN OVERBUILT MARKETS TABLE 2 Vacancy Assumptions Used in 1985]92 Calculations 1985 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
2.6 1.6 2 3 4 5 6 7 8 7 6.8
1986
1987
1988
1.6 1.2 2 3 4 5 6 7 8 7 6.8
1.2 2.3 3 4 5 6 7 8 7 6.8
2.3 3.9 5 6 7 8 7 6.8
1989
3.9 8 10 10 8 6 6 6.8
1990
9.4 13 15 13 10 7.5 5 5.5 5.5 6.8
1991
15.7 19 20.5 17 13 10 7.5 5.5 5 5 6 6.8
1992
23 24 23 21 19 17 15 12.5 10 7.5 5 4.5 4.5 5.5 6.8
Note. Data are those expected in June of the years on the horizontal to exist at the end of the years on the vertical.
vacancy rates.. The ‘‘expectations’’ for 1989]92 were developed in discussions with JLW staff ŽJLW was itself forecasting higher vacancy rates in 1989 and 1990.. Huge construction in process indicated a 7 percentage point rise in the Sydney vacancy rate between end1991 and end1992. The expected rate is assumed to remain in the 23% range throughout 1993]94, then to decline by 2 percentage points a year through 1998, at which point the decline accelerates to 2.5 points per year due to the recognition that some stock will be profitably converted to other uses. By 2003, the rate is 5%.14 After that it stays below the natural vacancy rate until actual real rents return to their equilibrium level in 2006.
14 The slight overshooting in vacancy rates after 2002 is needed to get rents to stabilize at their equilibrium level.
62
PATRIC H. HENDERSHOTT
1992 VrRC Calculations With the above described set of assumptions, the June 1992 VrRC ratio for office buildings in Sydney Australia is 0.435.15 Figure 3 shows the time paths by which the ratios of net cash flows to equilibrium net rents and value to replacement cost are expected to return to unity. Net cash flows in 1992 were just under half of equilibrium net rents, and they plunge to under one-fifth throughout the 1994]96 span. During the same period, real effective gross rents decline by 27%. Real gross rents then turn around gradually before jumping at a 9 to 13% annual rate during the 1999]2003 period. This rise, combined with the increase in occupancy, pushes net cash flows up rapidly. In contrast, the rise in the VrRC ratio is a virtual constant 10% per year over the 1994]99 period before dampening as the ratio dovetails into unity. Value rises immediately, in spite of the 60% further decline in real net cash flows, because the period of higher rents is coming closer. Of course, our assumptions are just that, so calculations are presented to indicate the sensitivity of the result to variations in the assumptions. First, we assume a more optimistic vacancy rate pattern: the rate declines by a point in 1993 and 2.5 points per year, reaching 7% in 1999 Žand then overshooting through 2003.. Not only is space occupied sooner, but rents adjust toward the equilibrium quicker. The result is a VrRC ratio of 0.565. On the other hand, a rental adjustment equation in which the response to the rent gap varies with the ratio of natural to actual vacancy rates fits the 1970]91 data about as well as the equation with a constant rent gap response implies a lower value. With this equation, rents initially decline further and then rise back to equilibrium less rapidly. As a result, the estimated VrRC ratio drops from 0.435 to 0.21. The ratio is not sensitive to the assumed real interest rate when feedback effects are incorporated. To illustrate, consider the impact of a decline in the real rate of interest from its unusually high June 1992 level. While the decline reduces below-equilibrium rents directly by lowering the equilibrium level, and thus raises the VrRC ratio, the time path along which actual rents rise to the lower equilibrium level is lower Žthe initial drop in rents is steeper., due to the reduced value of the equilibriumactual rent gap, and the present value of a given below-equilibrium rental shortfall is greater. When the 1992 ratio is recomputed for a 2 percentage point lower real rate, VrRC rises only from 0.435 to 0.45.
15 A comparable calculation of this ratio for U.S. office markets nationwide in early 1992 was 0.6 ŽHendershott and Kane, w5x.. A calculation for Melbourne Australia, using the Sydney rental adjustment equation, is 0.36, due to a sharper rise in vacancy rates and decline in real rents.
REAL ESTATE IN OVERBUILT MARKETS
63
FIG. 3. Adjustment of cash flows and value.
The reported VrRC ratios computed so far are for buildings earning market rents. However, in the 1980s Australia had lease contracts in which rents on existing buildings would rise with increases in market rents but not decline with decreases. To the extent that tenant rents have not ‘‘ratcheted’’ down in response to declines in market rents, the roughly 50% decline in real effective market rents observed between 1989 and 1992 was largely avoided. Because of their above-market leases, these buildings are worth more than their fundamental values. To calculate how much more buildings with above-market leases should be worth, real rents are adjusted upward. We consider buildings fully leased-up with tenants who obtained 10-year leases throughout the 1980s.16 Thus, one-tenth of the building ‘‘goes to market’’ each year over the next decade. Nominal rents on the nonmarket leases are constant Žreal rents fall by 5% per year.. By June 1999, all leases are at market. Under these assumptions, the replacement-cost ratio in 1992 for prime buildings, including both fundamental value and the premium for above-market leases, is 0.558. That is, fully enforceable above-market leases were worth another 0.12 of replacement cost.
16 The 10-year lease term is consistent with the computation of effective rents and has been the typical term for the Sydney office market. Just as the premium on bonds paying above-market coupons will be greater the longer the maturity of the bond, the value of above-market leases will be greater the longer the term of the lease.
64
PATRIC H. HENDERSHOTT
IV. FUNDAMENTAL, BOMA, AND JLW VALUES, 1985]92 Using our framework and appropriate assumptions, a time series of the valuerreplacement-cost ratio for Sydney office buildings is constructed for the 1985]92 period. In addition, comparable ‘‘market’’ ratios are computed from the BOMA and JLW data. Comparison of these ratios with that estimated using our method reveals substantial differences. The real interest rate, depreciation rate, and expense ratio are set at the 1992 values throughout the 1985]92 period. The market vacancy rate forecasts were listed in Table 2. For the June 1985 through June 1988 period, the year-end and next year vacancy rates are assumed to be accurately forecast; the vacancy rate is then increased a percentage point a year to 8% and then reduced over 2 years to the natural rate. For 1989]91, a sharp increase in the vacancy rate Žbased on discussions with JLW. is followed by a reversal, including a brief time below the natural rate. Column 1 of Table 3 contains the calculated Sydney VrRC ratios for the 1985]92 period. The ratio is roughly constant at 1.0 over the 1985]89 period Žthe 1989 ratio is no higher than the 1985 ratio, in spite of the 50% greater real effective rent level in 1989, because of the substantially higher vacancy rates}and thus real rent declines}forecasted in 1989 than in 1985.. Subsequent sharp jumps in vacancy rates and plunging real effective rents lower the ratio to 0.435 in 1992. The 1992 premium for the value of above-market leases is also listed in the table. Two sources publish market data from which VrRC ratios can be computed and compared with the fundamental value ratio adjusted for
TABLE 3 VrRC Ratios for Sydney, 1985]92
June 1985 1986 1987 1988 1989 1990 1991 1992 Premium leases Ž1992.
Ž1.
Ž2.
Ž3.
VrRC
BOMA
1.01 1.02 1.02 1.01 0.98 0.86 0.68 0.43
1.01 1.05 1.13 1.32 1.54 1.53 1.25 1.01
0.12
JLW
Ž4. wŽ2.rŽ1.x Relative BOMA
Ž5. wŽ3.rŽ1.x Relative JLW
1.01 1.04 1.14 1.46 1.69 1.61 1.17 0.76
1 1.03 1.11 1.31 1.57 1.72 1.69 1.94
1 1.02 1.12 1.44 1.72 1.80 1.58 1.47
REAL ESTATE IN OVERBUILT MARKETS
65
above-market leases. A comparable BOMArRussell series comes from a two-step procedure. First, the appreciation component of the BOMArRussell office return series is cumulated to generate a nominal appraised value series. Second, to obtain a real constant-quality measure, this series is divided by the deflator for private consumption Žscaled to unity in June 1985. and then blown up by 1.25% per year to allow for depreciation in the replacement-cost denominator. ŽThe 1.25% is less than the general 2.5% depreciation rate because new buildings are continuously added to the BOMA data base.. The BOMArRussell series is set equal to 1.01, the fundamental valuerRC ratio, in June 1985, which is consistent with our assumption that rents equaled their equilibrium values in 1986. These values are shown in column 2 of Table 3. For the JLW calculation, we begin with their nominal capital value indicator and adjust it as in step two, using a 2.5% depreciation factor. The resultant VrRC estimates, including the assumed 1.01 1985 value, are listed in column 3. The BOMA and JLW data increase sharply between June 1985 and June 1989, 53 and 68%, respectively, and then decline abruptly between June 1989 and June 1992, 34 ŽBOMA. and 57% ŽJLW.. The smaller BOMA movements are consistent with appraised values lagging market values.17 Both series, but especially that based on BOMA data, are substantially above the fundamental value in June 1992. Two possible explanations exist for the surge in the BOMA and JLW series in the second half of the 1980s. First, the real required return on real estate Žthe risk premium in our model. may have declined. Second, a speculative bubble Žincrease of market relative to fundamental value. may have formed. One indicator of a speculative bubble, given the constancy of the fundamental value series, would be an independent estimate of a sharp increase in real land prices.18 Such an estimate exists. The Valuer-General of New South Wales computed a 350% real increase in land value per square meter of office space in the Sydney central financial district between 1986 and 1989.19 If land were only 5% of the total value of offices constructed in 1986, the real land increase would account for virtually all of the computed increases in the BOMA and JLW ratios.
17
Like the Russell]NCREIF data in the United States, the BOMA]Russell data in Australia are largely based on appraised values. The JLW data are JLW’s estimates of market values. 18 In our valuation model, land is valued together with the structure; the rents that the two can earn jointly determines the value of the total property. 19 Land value data are contained in various reports; the latest are in Valuer-General Žw12x, Table 18, p. 21.. These data pertain to the middle of the year and are supposed to represent market values based on the comparable sales method. The Valuer-General’s estimates, which are the basis for property tax assessments, may be challenged legally.
66
PATRIC H. HENDERSHOTT
Because land values may be mismeasured in the RC measure, we compute the ratios of the BOMArRussell and JLW based ratios to our fundamental value Žadjusted for leases. ratio, thereby eliminating measurement error in RC. Ratios of ratios, assuming 0.03, 0.06, and 0.09 in 1990]92, respectively, to be the ratio of enforceable above-market leases to replacement cost, are reported in columns 4 and 5. As can be seen, JLW’s 1992 ratio is almost 50% greater than our estimate of fundamental value plus the value of above-market leases, and the ratio derived from BOMA’s 1992 data is almost double our estimate.20 These multiples suggest that the late 1980s office market bubble has not yet burst. However, data from the New South Wales Valuer-General suggest that the bubble has popped; real office market land prices have returned to their 1986 values. Moreover, any decline in the real estate risk premium during the 1985]89 period has almost certainly also been reversed. The implication is that market value should be in line with fundamental value Žplus above-market leases.. We conclude that the JLW, and especially BOMA, Sydney office market data overstated market value significantly in June 1992.21 V. SUMMARY Real estate markets are periodically plagued by excess supply, rent concessions, and few arms-length transactions. During such periods, valuation is problematic. The model presented here expresses fundamental value as the difference between replacement cost and the present value of expected below-equilibrium rents. Equilibrium rents are specified and real rental adjustment equations are estimated using Sydney office market data during the 1970]92 period. The percentage change in real effective office rents is related to gaps between both the natural and actual vacancy rates and equilibrium and actual real rental rates. Inclusion of the rent gap guarantees that actual rents will eventually return to their equilibrium level Žand more than doubles the equation’s explanatory power.. With this relationship, only the market participants’ forecast of vacancy rates until the market returns to equilibrium is needed to estimate value. Such a forecast was obtained from the Sydney JLW consulting office. The estimated fundamental-valuerreplacement-cost ratio for the Sydney office market was about 1.0 throughout the mid-1985 to mid-1989 period. The subsequent surge in vacancies from 3 to 23% halved effective rents and lowered the ratio to 0.43 by mid-1992. For prime pre-1989 20 The JLW overestimate can be largely explained by their mismeasurement of the decline in effective real rents during the 1989]92 period. 21 The JLW difference for Melbourne is comparable; the BOMA ratio is over three times greater than our estimate of fundamental value plus the value of above-market leases.
REAL ESTATE IN OVERBUILT MARKETS
67
properties with enforceable ‘‘rachet’’ clauses, real rents have only begun their plunge. The above-market rents these buildings are earning could be worth another 0.12 of replacement cost. Ratios developed from the capital return component of the BOMA return series and from the JLW capital value series indicate about a 50% increase in the ratio in the second half of the 1980s. Even with a sharp decline since then, these ratios imply values 50 to 100% greater than our estimate of fundamental value plus enforceable above-market leases. While evidence on Sydney land values and of an office market construction boom supports the conjecture of a speculative bubble forming in the second half of the 1980s, the same evidence implies that the bubble has burst and that the BOMA data especially significantly overstate market value.
REFERENCES 1. A. Eubank and C. F. Sirmans, The price adjustment mechanism for rental housing, Quarterly Journal of Economics, 93, 163]83 Ž1979.. 2. R. Flood and R. Hodrick, On testing for speculative bubbles, Journal of Economic Perspecti¨ es, 4, 85]101 Ž1990.. 3. P. Hendershott, J. Follain, and D. Ling, Effects on real estate, in ‘‘Tax Reform and the U.S. Economy’’ ŽJoseph A. Pechman, Ed.., The Brookings Institution, Washington, DC Ž1987.. 4. P. Hendershott and E. Kane, The causes and consequences of the 1980s commercial real estate boom, Journal of Applied Corporate Finance, 61]70 Ž1992.. 5. P. Hendershott and E. Kane, ‘‘Office Market Values During the Past Decade: How Distorted Have Appraisals Been?,’’ Real Estate Economics. 6. Patric H. Hendershott and Yunhi Won, Introducing risky housing and tenure choice into a portfolio-based general equilibrium model, Journal of Public Economics, 48, 293]316 Ž1992.. 7. JLW Research and Consulting, ‘‘The Implications of Leasing Incentives for Estimating Rents and Capital Values,’’ Property Research Paper, Sydney, July Ž1992.. 8. D. Ling, Real estate values, federal income taxation, and the importance of local market conditions,’’ AREUEA Journal, 20, 125]139 Ž1992.. 9. K. Rosen and L. Smith, The price adjustment process for rental housing and the natural vacancy rate, American Economic Re¨ iew, 73, 779]86 Ž1983.. 10. A. Shleifer and L. H. Summers, The noise trader approach to finance, Journal of Economic Perspecti¨ es, 4, 19]34 Ž1990.. 11. J. Stiglitz, Symposium on bubbles, Journal of Economic Perspecti¨ es, 4, 13]18 Ž1990.. 12. Valuer-General, ‘‘New South Wales Real Estate Market 1992,’’ Valuer-General’s Department, Sydney, New South Wales Ž1992.. 13. W. Wheaton and R. Torto, Office rent indices and their behavior over time, Journal of Urban Economics, 35, 121]139 Ž1994.. 14. W. Wheaton and R. Torto, Vacancy rates and the future of office rents, AREUEA Journal, 16, 430]437 Ž1988..
39, 51]67 Ž1996.
0003
Rental Adjustment and Valuation in Overbuilt Markets: Evidence from the Sydney Office MarketU PATRIC H. HENDERSHOTT Max M. Fisher College of Business, The Ohio State Uni¨ ersity and NBER, 1775 College Road, Columbus, Ohio 43210 Received May 9, 1994; revised August 29, 1994
In equilibrium, real estate will be valued at Žapproximately. replacement cost. When value exceeds replacement cost in a market, new construction surges, raising vacancies and lowering rents and thus value. When replacement cost exceeds value, replacement construction is deferred, lowering vacancies and raising rents and value. Moreover, expectations of the construction, vacancy, and rent responses imply that existing real estate values will be only modestly sensitive to changes in required real after-tax returns caused by real interest rate or tax changes.1 Because higher long-term real interest rates lower real estate values, construction declines and rental income on existing properties rises. That is, the rental cash flows on existing real estate are more analogous to those on a sluggishly adjusting adjustable-rate loan than to the fixed coupons on a long-term bond; thus the percentage real estate value decline would be far less than that of a long-term bond. U
I thank BOMA for supplying me with data, Bond University for making facilities available to me while the first draft of the paper was being written, and Frank Russell Australia for financial support. The cooperation of the Sydney JLW Research and Consultancy Office, which supplied me with real rental data Žincluding rent concessions. and vacancy rate projections of market participants, was crucial to the success of this project. Individuals supplying useful input include Adrian Harrington of BOMA, David Dickinson, N. H. Seek, and S. G. Slim of the Sydney JLW Research and Consultancy Office, John Bowers, Ralph Jackson, and David MacKenzie of Frank Russell, and Barry Reece of the University of NSW. I also thank two anonymous reviewers and the journal editor. Earlier versions of this paper were presented at the Univ of ConnrAREUEArFannie Mae 1993 International Real Estate Conference, Mystic, CT, October 1993 and the first Global Finance Conference in Monterey, CA, April 1994. 1 Hendershott, Follain, and Ling w3x make this point in their analysis of the impact of the 1986 tax act on real estate values. Ling w8x extends the argument in his analysis of the impact of capital gains tax rate changes. 51 0094-1190r96 $12.00 Copyright Q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.
52
PATRIC H. HENDERSHOTT
As a result of the construction response mechanism, deviations of real estate values greater than five percent from replacement cost would be rare in relatively stable markets, i.e., those without substantial excess supply or demand. Unfortunately, stable well-functioning is not the way one would describe office markets in many parts of the western world. In London, Tokyo, most national capitals of Scandinavia, most state capitals of Australia, and many large U.S. cities, vacancy rates of 20 to 30% have been common, real effective rents have plunged, and values are almost certainly far below replacement cost. In the transforming socialist economies of Central and Eastern Europe, the profit-motivated construction equilibrating mechanism described above has not existed for at least a half a century. For the value of any type real estate in any locale there to be near replacement cost would be pure happenstance. This paper illustrates a method for valuing real estate in overbuilt Žor underbuilt. markets with office building data from Sydney, Australia. The method is presented in Section I, where fundamental value is expressed as replacement cost less the present value of expected below-equilibrium real rental income. Underpinnings for the calculation of below-equilibrium real rents are given in Section II; equilibrium real rents are specified and real rental adjustment equations are estimated using data from the 1970]92 period. The percentage change in real effective rents is related to gaps between both the natural and actual vacancy rates and equilibrium and actual real rental rates. Expectations of future equilibrium real rents Žincluding real discount rates. and vacancy rates thus provide a future rental income stream from which the present value of expected belowequilibrium rental income, and thus fundamental value, can be computed. The valuerreplacement-cost ratio is calculated for Sydney in Section III, and its sensitivity to a number of assumptions is indicated. Most importantly, the 1992 ratio is not sensitive to real interest rate Žand thus risk premium. assumptions due to offsetting adjustments in below equilibrium rents and the discount rate. Also, the value of above-market leases, which likely existed on Sydney buildings leased up in the early 1990s, is computed. In the absence of price ‘‘bubbles,’’ price should equal fundamental value plus the value of above-market leases. 2 In Section IV, estimates of the ratio of fundamental values plus above-market leases to replacement cost are compared and contrasted with ratios derived from series commonly reported in two prominent Australian real estate sources, JLW and
See Flood and Hodrick w2x, Shleifer and Summers w10x and Stiglitz w11x for useful discussions of the literature on asset market bubbles. While the construction response will not guarantee the absence of bubbles, it will clearly limit their duration. 2
53
REAL ESTATE IN OVERBUILT MARKETS
BOMA.3 During the 1990]92 period, the JLW and BOMA series were 50 to 100% greater than fundamental value Žplus the value of enforceable above-market leases.. An important issue is whether the reported series are misrepresentations of market value or reflect a speculative bubble. While independent evidence on Sydney land values is offered for the development of a speculative bubble in the second half of the 1980s, the same evidence implies the bubble burst in the early 1990s. Further, the surge in Sydney office market construction in the late 1980s, a reasonable builder response to a prices exceeding fundamental value, almost guarantees the bursting of such a bubble in the early 1990s. That is, the 1992 BOMA data Žand the JLW data to a lesser extent. appear to overstate market value significantly. 4 I. THE BASIC VALUATION MODEL The fundamental value of a property in period t Ž Vt . is the present value of expected future net operating income. Assuming, for simplicity, that the expected ratio of operating expenses to replacement cost Žoper. and one-period real risky financing rates Ž r . are constant through time, value at time t for a tax-exempt investor is
Vt s
`
Ý js1
g tqj Ž 1 y ¨ tqj . y exp RCt Ž 1 y d .
Ž 1 q rtU .
j
j
,
Ž 1.
where g is the expected real gross rental rate per replacement-cost dollar of occupied space, ¨ is the expected vacancy rate, and RC, the current replacement cost of built space, is expected to decline at the depreciation rate, d.5 3 The Sydney Jones]Lang]Wootten ŽJLW. Research and Consultancy Office reports effective rent, vacancy rate, and capital value series for most Australian capital cities. The Building and Operating Management Association ŽBOMA. reports cash flow and appreciation returns for different property types in most capital series. The BOMA series are analogous to those produced by the Frank Russell company in the United States Žthe appreciation return component is based on appraisals, not transactions .. 4 Some have argued that appraisers are reluctant to revalue properties sharply over short periods. This tendency could be exaggerated for downward adjustments due to the incentives of investment managers to maintain the values upon which their percentage fees are based. 5 Technically, the zero-coupon rate Žplus an appropriate risk premium. for each maturity should be used to discount each period’s cash flow. When the yield curve is steeply sloped and the cash flows are markedly uneven, using different-period discount rates can greatly affect valuation. Similarly, the expected inflation rate between the current period and when the cash flow will be received should be used to compute each cash flow.
54
PATRIC H. HENDERSHOTT
In equilibrium, ¨ tqj will equal ¨ U for all j, gross rents will equal real financing costs, economic depreciation, and operating expenses, and value will equal replacement cost. More specifically, if g Ž 1 y ¨ U . s r U q d q exp s g U
Ž 2.
and ¨ s ¨ U Žthe natural vacancy rate., then value will equal replacement cost. The validity of Ž1. can be verified by setting all ¨ tqj s ¨ U , all g tqj s g Ut , and substituting Ž2. into Ž1.. The determinants of the VrRC ratio are of particular interest to us. To place these determinants in better focus, we add and subtract `
Ý
Ž rt q d . RCt Ž 1 y d . Ž 1 q rt .
js1
jy1
s RCt
j
from the right hand side of Eq. Ž1.. Combining terms and dividing by RC, we see that the VrRC ratio is reduced from unity to the extent that expected future below-equilibrium rents ŽBERI. exist: Vt RCt
s1y
BERI t RCt
,
Ž 3.
where BERI t RCt
N
s
Ý js1
j Ž rtU q d q exp. y g tqj Ž 1 y ¨ tqj . Ž 1 y d .
Ž 1 q rtU .
j
.
Ž 4.
The model has been set up in terms of tax-exempt investors. A valuation model for taxable investors can be obtained by introducing taxes into Eqs. Ž1. and Ž2..6 Regarding Eq. Ž1., the cash flows would be taxed, an after-tax discount rate would be used, and a term to reflect the value of tax depreciation would be added. In addition, equilibrium net rents in Eq. Ž2. would depend upon individual tax rates and tax depreciation. However, in an overall equilibrium, net rents for all investor classes will be equal; the required after-tax returns and portfolio shares invested in real estate will differ across investor classes, with more heavily taxed investors having smaller portfolio shares and thus smaller risk premiums.7 In disequilibrium, properties will shift toward whatever investors have the larger valuation. The required returns of these investors will rise and returns for See Hendershott and Kane w5x for full presentation of such a model. For analysis of a portfolio model in which investors in different tax brackets and with different risk-aversion parameters}and thus different required returns}hold different portfolio shares, see Hendershott and Won w6x. 6 7
REAL ESTATE IN OVERBUILT MARKETS
55
the selling investors will fall until all investors stand again in equilibrium, with properties valued equally by all investor classes. That is, appropriately specified and parameterized, taxable and tax-exempt models will yield the same valuations. The keys to valuation are first specifying equilibrium rents Žthe real interest and depreciation rates and the operating expense ratio in the tax-exempt model. and second determining both how far the rental rate Ž g adjusted for vacancies. is currently below the equilibrium rate and the time path along which the rental rate will return to equilibrium. The latter involves projecting the time path of vacancy rates and the real gross rental share. While we have estimates of market participants’ expectations of future vacancy rates, expected real gross rents are based upon an estimated real gross rental adjustment equation. This is the topic of the next section. II. EXPLANATION OF SYDNEY REAL EFFECTIVE RENTS A long tradition exists of relating the percentage change in real effective rents to the difference between the ‘‘natural’’ and observed vacancy rates w1, 9, 14x: D g tqjrg tqjy1 s lŽ ¨ U y ¨ tqjy1 . .
Ž 5.
As popular as this reduced-form relation has been, it suffers from three problems. First, the equation provides no relationship between the le¨ els of actual and equilibrium rents, which must be equal in equilibrium. Second, the adjustment equation requires substantial overshooting of the natural rate in response to shocks other than a change in the equilibrium rental rate.8 Third, the equation cannot hold simultaneously for leases of different terms. To illustrate these problems, consider a market initially in equilibrium with gross rents equal to equilibrium rents, the vacancy rate equal to the natural rate, and value equal to replacement cost. A negative demand or positive supply shock Žexcess space unexpectedly becomes available. increases vacancies and decreases rents Žand thus values.. The contraction in construction resulting from value declines turns vacancies around, but rents continue to fall until the vacancy rate is back to the natural rate. The market must then go through a reverse adjustment with rents rising, first with falling vacancies and later with rising vacancies, until they return to equilibrium, vacancies return to the natural rate, and value again equals Wheaton and Torto w13x discuss the microfoundations of the adjustment equation and estimate a less constrained form for numerous U.S. cities on data from the 1979]91 period. They do not impose an equilibrium rent relationship. 8
56
PATRIC H. HENDERSHOTT
replacement cost. To create the force returning rents to equilibrium, the rental adjustment equation necessarily entails the vacancy rate substantially overshooting the natural rate. The vacancy rate will decline below the natural rate because value is less than replacement cost Žnet construction is negative. when the actual vacancy rate first returns to the natural rate. Moreover, even if the rental adjustment equation characterized oneperiod leases, it cannot hold for longer term leases. With rational expectations, the rent on multiperiod leases will be an average of the expected future rents on one-period leases, just as longer term bond rates are averages of future one-period bond rates. Thus expectations that oneperiod rents will rise in the future will turn rents on multiperiod leases upward before the vacancy rate reaches its natural level.9 Logically, Eq. Ž5. cannot hold simultaneously for leases of different term. To allow a more general adjustment path with pleasing long-run properties, real effective rents are specified as adjusting to gaps between both the natural and actual vacancy rates Ž ¨ U and ¨ y1 . and equilibrium and actual gross rents Ž g U and gy1 .: D g tqjrg tqjy1 s lŽ ¨ U y ¨ tqjy1 . q b Ž g Utqj y g tqjy1 . .
Ž 6.
With this equation, vacancy rates do not have to overshoot following a supply shock. After high vacancy rates have dragged rents significantly below equilibrium, the known eventual return to equilibrium acts as a force causing real rents to rise, even when the vacancy rate is still above the natural rate. Moreover, with ¨ s ¨ U , real rents will adjust to their equilibrium value. The Data Rental adjustment equations are estimated using annual data from Sydney over the 1970]92 period.10 The natural vacancy rate is treated as a constant over time.11 In Ž5. or Ž6., ¨ tqjy1 is included as a regressor and the constant term is l¨ U . Thus the natural rate can be computed after the 9
In a cross-sectional study of individual leases, the term of the lease can be entered as a regressor to capture the differential impact of future expected values of one-term leases on the value of longer-term leases with different terms. In a pooled time-series cross-sectional study, the impact of term should be allowed to vary over time because the term structure of leases varies over time. 10 All data were generously supplied by the Sydney JLW Research and Consultancy office w7x. 11 Few degrees of freedom prevent serious testing of a nonconstant natural vacancy rate. It is worth noting, though, that reported evidence seeming to suggest large time variation in natural vacancy rates may simply reflect continuing market disturbances ŽHendershott and Kane, w5x..
REAL ESTATE IN OVERBUILT MARKETS
57
estimation as the ratio of the constant term to the coefficient on the lagged vacancy rate Ž l.. Real effective rents require the downward adjustment of real face rents for both free rent periods and tenant improvements. Conveniently for us, JLW w7x has collected data on both of these terms for Sydney office markets over the 1970]91 period. According to their data, rent incentives increased sharply over the 1989]91 period, rising from less than 4 months equivalent free months rent Žincluding adjustment of improvements. on a ten-year lease to almost 23 free months. JLW also compute effective real rents by discountingramortizing real cash flows with a real interest rate. We have adjusted JLW’s methodology in only one respect; they amortize the present value of the rent incentives over the life of the building whereas the correct amortization period is the length of the lease. With this adjustment, the amortized value of the rent incentives jumped from 6% of real face rent in 1989 to 41% in 1992.12 The percentage change in the resultant real effective rent is the dependent variable in our analysis. The equilibrium real gross rental rate equals the real risk-free rate plus a risk premium, the depreciation rate, and the expense ratio. An approximate Australian risk-free real rate was computed as Ž1 q r f .rŽ1 q p . y 1 for semi-annual intervals over the December 1969 to June 1992 period, using the 10-year Treasury rate as the risk-free rate Ž r f . and a three-period average of annualized percentage changes in the Deflator for Private Final Consumption Expenditures as the expected inflation proxy Žp .. The series actually employed is a smoothed version of the calculated series because longer term real interest rates should move relatively smoothly. The smoothed real risk-free rate drops from 0.02 in 1970 to zero in early 1971, stays there until the middle of 1974, abruptly plunges to y0.04, and remains there through the middle of 1977. The series then rises linearly to 0.06 in June 1982, where it stays through June 1992. The risk premium and depreciation rates are set at 0.035 and 0.025, respectively, throughout the period. BOMA data indicate that the ratio of outgoings Žoperating expenses. to gross rates has ranged from 0.285 to 0.335; oper is set equal to 0.05, giving a gross rental rate of 0.17 and an outgoings ratio of 0.294 Ž0.05r0.17.. Thus the equilibrium rental rate declines from 0.13 in 1970 to 0.07 in the 1975]77 period and rises to 0.17 throughout the 1982]92 span. To convert this gross rental rate series to a current Ž1992. dollar rent per square meter, the series is multiplied by the real rent level in a year in which equilibrium and actual rents appear to have been equal. Given the steadiness of both actual and equilibrium rents 12
Their methodology understates the increase in value of rent incentives and thus the decrease in effective rents between 1989 and 1992. Rather than a 48% real decline, they compute a 26% decline.
58
PATRIC H. HENDERSHOTT
during the 1983]85 period, real effective rents are assumed to have equaled equilibrium rents in June 1986. This equilibrium rent series and the JLW real effective rent series are plotted in Fig. 1. A clear relationship is evident, with the JLW series generally lagging the equilibrium series. The last key input is the vacancy rate. This series began and ended the 1970s at about 3% with a sharp increase and reversal in between Žthe peak was 13% at the end of 1976.. Real effective rents followed the vacancy rate with about a 2 year lag. The vacancy rate stayed below 5% throughout the 1980s and was 3% in the middle of 1989; during the second half of the decade, real effective rents rose by 50%. Since then a plethora of completions and negative absorption has driven vacancy to 23% and halved real effective rents. The Estimates Table 1 reports some estimates of the rental adjustment equation. Coefficient estimates and standard errors Žin parentheses ., the equation R 2 Žnot adjusted for degrees of freedom., and the implied natural vacancy rate are reported. The first estimate is of the naive model in which rents adjust solely to the difference between the natural and actual vacancy rates. As is seen, the equation explains less than a third of the variance. The implied natural vacancy rate is 6.0%. Adding the equilibrium rent gap more than doubles the explanatory power. While the explanatory power is reasonable, the equation misses a significant portion of the broad swings in real effective rents. More specifically, between 1978 and 1989 ŽJune. real effective rents per square meter rose by $513, $375 of which was lost by June 1992. The equation explains only 72 and 38%, respectively, of the increase and reversal.
FIG. 1. Actual and equilibrium rents.
59
REAL ESTATE IN OVERBUILT MARKETS
TABLE 1 Explanation of the Percentage Change in Real Effective Sydney Office Market Rents, 1971]92 Equ.
Constant
¨ y1
1.1
.1125
1.2
.1121
1.3
.1159
y.0189 Ž.0062. y.0176 Ž.0047. y.0171 Ž.0046.
g U y gy1
2.57 Ž0.60. 2.26 Ž0.64.
D¨
y.0085 Ž.0069.
R2
¨U
.317
6.0
.650
6.4
.679
6.8
Note. The data are from JLW. Vacancy rates are in percentage points; rental rates are in decimals. Standard errors are in parentheses. See the text for exact definition of the variables.
To better explain the sharp fall in real rents during the June 1989]June 1992 period, the forward change in the vacancy rate Žthe December to December change explaining the June to June change in rents. has been added. Knowing that substantial capacity was coming on line should have directly reduced real effective rents. While adding the forward vacancy-rate change does not increase the explanatory power much Žthe t-ratio on the variable is barely above one., the equation is far better able to explain the pre-1989 rise in real rents and especially the subsequent decline. Rather than the 72 and 38% just mentioned, 85% of the increase and 63% of the decrease are explained. The estimate of the natural vacancy rate rises slightly to 6.8%. Figure 2 plots the level of real effective rents and the level predicted dynamically Žthe initial level moved forward based on the predicted changes.. Because the annual errors in predicted changes are largely independent, the predicted level tracks the actual level reasonably well. Figure 2 also includes forecasted rent levels for the 1993]2003 period, assuming the Sydney vacancy rate is 23% through December 1994 and then declines Žsee below.. The equation predicts another $120 Ž27%. decline in real effective rents before an upturn in 1998.13 III. VALUE r REPLACEMENT-COST RATIOS IN JUNE 1992 In this section, the model is fully parameterized and VrRC ratios are computed for Sydney in June 1992 under the base assumptions using each of the real rental adjustment equations. The sensitivity of the calculated 13 According to recent JLW data, real effective rents declined by 30% between June 1992 and June 1993.
60
PATRIC H. HENDERSHOTT
FIG. 2. Actual and predicted real effective rents.
VrRC ratio to each of a more optimistic vacancy rate decline, alterations in estimated rental-adjustment coefficients, and a decrease in real interest rates is illustrated. Last, the value of long-lived enforceable leases, which would result in slower declines in real rents after 1989, is also computed. Model Parameterization The VrRC ratio is unity less the present value of below-equilibrium future cash flows divided by replacement cost. As can be seen from Eqs. Ž2. and Ž4., the variables to be specified are: the real risky discount rate Ž r ., the depreciation rate Ž d ., the operating expense ratio Žoper., the natural vacancy rate Ž ¨ U ., and the time paths of future vacancy and gross rental rates. Moreover, from Eq. Ž6., the time path of future gross rental rates is fully specified by the other variables. With constant future values of r, d, ¨ U , and oper, only a future time path of vacancy rates is needed for the calculations. Most of the June 1992 values have already been specified: r s 0.095, d s 0.025, and oper s 0.05, giving a gross equilibrium rental rate of 0.17. From the estimation, ¨ U s 0.068. With the assumption of equality between actual and equilibrium rents in June 1986, the actual gross rental rate in June 1992 is 0.117. The last key assumption is the future time path of vacancy rates expected by market participants. Table 2 lists assumed future vacancy rates in Sydney as of year-end over the 1985]92 period. Each column contains the rate expected during June of the year heading the column to exist at the end of the years in the far left column Žthe diagonal contains observed
61
REAL ESTATE IN OVERBUILT MARKETS TABLE 2 Vacancy Assumptions Used in 1985]92 Calculations 1985 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
2.6 1.6 2 3 4 5 6 7 8 7 6.8
1986
1987
1988
1.6 1.2 2 3 4 5 6 7 8 7 6.8
1.2 2.3 3 4 5 6 7 8 7 6.8
2.3 3.9 5 6 7 8 7 6.8
1989
3.9 8 10 10 8 6 6 6.8
1990
9.4 13 15 13 10 7.5 5 5.5 5.5 6.8
1991
15.7 19 20.5 17 13 10 7.5 5.5 5 5 6 6.8
1992
23 24 23 21 19 17 15 12.5 10 7.5 5 4.5 4.5 5.5 6.8
Note. Data are those expected in June of the years on the horizontal to exist at the end of the years on the vertical.
vacancy rates.. The ‘‘expectations’’ for 1989]92 were developed in discussions with JLW staff ŽJLW was itself forecasting higher vacancy rates in 1989 and 1990.. Huge construction in process indicated a 7 percentage point rise in the Sydney vacancy rate between end1991 and end1992. The expected rate is assumed to remain in the 23% range throughout 1993]94, then to decline by 2 percentage points a year through 1998, at which point the decline accelerates to 2.5 points per year due to the recognition that some stock will be profitably converted to other uses. By 2003, the rate is 5%.14 After that it stays below the natural vacancy rate until actual real rents return to their equilibrium level in 2006.
14 The slight overshooting in vacancy rates after 2002 is needed to get rents to stabilize at their equilibrium level.
62
PATRIC H. HENDERSHOTT
1992 VrRC Calculations With the above described set of assumptions, the June 1992 VrRC ratio for office buildings in Sydney Australia is 0.435.15 Figure 3 shows the time paths by which the ratios of net cash flows to equilibrium net rents and value to replacement cost are expected to return to unity. Net cash flows in 1992 were just under half of equilibrium net rents, and they plunge to under one-fifth throughout the 1994]96 span. During the same period, real effective gross rents decline by 27%. Real gross rents then turn around gradually before jumping at a 9 to 13% annual rate during the 1999]2003 period. This rise, combined with the increase in occupancy, pushes net cash flows up rapidly. In contrast, the rise in the VrRC ratio is a virtual constant 10% per year over the 1994]99 period before dampening as the ratio dovetails into unity. Value rises immediately, in spite of the 60% further decline in real net cash flows, because the period of higher rents is coming closer. Of course, our assumptions are just that, so calculations are presented to indicate the sensitivity of the result to variations in the assumptions. First, we assume a more optimistic vacancy rate pattern: the rate declines by a point in 1993 and 2.5 points per year, reaching 7% in 1999 Žand then overshooting through 2003.. Not only is space occupied sooner, but rents adjust toward the equilibrium quicker. The result is a VrRC ratio of 0.565. On the other hand, a rental adjustment equation in which the response to the rent gap varies with the ratio of natural to actual vacancy rates fits the 1970]91 data about as well as the equation with a constant rent gap response implies a lower value. With this equation, rents initially decline further and then rise back to equilibrium less rapidly. As a result, the estimated VrRC ratio drops from 0.435 to 0.21. The ratio is not sensitive to the assumed real interest rate when feedback effects are incorporated. To illustrate, consider the impact of a decline in the real rate of interest from its unusually high June 1992 level. While the decline reduces below-equilibrium rents directly by lowering the equilibrium level, and thus raises the VrRC ratio, the time path along which actual rents rise to the lower equilibrium level is lower Žthe initial drop in rents is steeper., due to the reduced value of the equilibriumactual rent gap, and the present value of a given below-equilibrium rental shortfall is greater. When the 1992 ratio is recomputed for a 2 percentage point lower real rate, VrRC rises only from 0.435 to 0.45.
15 A comparable calculation of this ratio for U.S. office markets nationwide in early 1992 was 0.6 ŽHendershott and Kane, w5x.. A calculation for Melbourne Australia, using the Sydney rental adjustment equation, is 0.36, due to a sharper rise in vacancy rates and decline in real rents.
REAL ESTATE IN OVERBUILT MARKETS
63
FIG. 3. Adjustment of cash flows and value.
The reported VrRC ratios computed so far are for buildings earning market rents. However, in the 1980s Australia had lease contracts in which rents on existing buildings would rise with increases in market rents but not decline with decreases. To the extent that tenant rents have not ‘‘ratcheted’’ down in response to declines in market rents, the roughly 50% decline in real effective market rents observed between 1989 and 1992 was largely avoided. Because of their above-market leases, these buildings are worth more than their fundamental values. To calculate how much more buildings with above-market leases should be worth, real rents are adjusted upward. We consider buildings fully leased-up with tenants who obtained 10-year leases throughout the 1980s.16 Thus, one-tenth of the building ‘‘goes to market’’ each year over the next decade. Nominal rents on the nonmarket leases are constant Žreal rents fall by 5% per year.. By June 1999, all leases are at market. Under these assumptions, the replacement-cost ratio in 1992 for prime buildings, including both fundamental value and the premium for above-market leases, is 0.558. That is, fully enforceable above-market leases were worth another 0.12 of replacement cost.
16 The 10-year lease term is consistent with the computation of effective rents and has been the typical term for the Sydney office market. Just as the premium on bonds paying above-market coupons will be greater the longer the maturity of the bond, the value of above-market leases will be greater the longer the term of the lease.
64
PATRIC H. HENDERSHOTT
IV. FUNDAMENTAL, BOMA, AND JLW VALUES, 1985]92 Using our framework and appropriate assumptions, a time series of the valuerreplacement-cost ratio for Sydney office buildings is constructed for the 1985]92 period. In addition, comparable ‘‘market’’ ratios are computed from the BOMA and JLW data. Comparison of these ratios with that estimated using our method reveals substantial differences. The real interest rate, depreciation rate, and expense ratio are set at the 1992 values throughout the 1985]92 period. The market vacancy rate forecasts were listed in Table 2. For the June 1985 through June 1988 period, the year-end and next year vacancy rates are assumed to be accurately forecast; the vacancy rate is then increased a percentage point a year to 8% and then reduced over 2 years to the natural rate. For 1989]91, a sharp increase in the vacancy rate Žbased on discussions with JLW. is followed by a reversal, including a brief time below the natural rate. Column 1 of Table 3 contains the calculated Sydney VrRC ratios for the 1985]92 period. The ratio is roughly constant at 1.0 over the 1985]89 period Žthe 1989 ratio is no higher than the 1985 ratio, in spite of the 50% greater real effective rent level in 1989, because of the substantially higher vacancy rates}and thus real rent declines}forecasted in 1989 than in 1985.. Subsequent sharp jumps in vacancy rates and plunging real effective rents lower the ratio to 0.435 in 1992. The 1992 premium for the value of above-market leases is also listed in the table. Two sources publish market data from which VrRC ratios can be computed and compared with the fundamental value ratio adjusted for
TABLE 3 VrRC Ratios for Sydney, 1985]92
June 1985 1986 1987 1988 1989 1990 1991 1992 Premium leases Ž1992.
Ž1.
Ž2.
Ž3.
VrRC
BOMA
1.01 1.02 1.02 1.01 0.98 0.86 0.68 0.43
1.01 1.05 1.13 1.32 1.54 1.53 1.25 1.01
0.12
JLW
Ž4. wŽ2.rŽ1.x Relative BOMA
Ž5. wŽ3.rŽ1.x Relative JLW
1.01 1.04 1.14 1.46 1.69 1.61 1.17 0.76
1 1.03 1.11 1.31 1.57 1.72 1.69 1.94
1 1.02 1.12 1.44 1.72 1.80 1.58 1.47
REAL ESTATE IN OVERBUILT MARKETS
65
above-market leases. A comparable BOMArRussell series comes from a two-step procedure. First, the appreciation component of the BOMArRussell office return series is cumulated to generate a nominal appraised value series. Second, to obtain a real constant-quality measure, this series is divided by the deflator for private consumption Žscaled to unity in June 1985. and then blown up by 1.25% per year to allow for depreciation in the replacement-cost denominator. ŽThe 1.25% is less than the general 2.5% depreciation rate because new buildings are continuously added to the BOMA data base.. The BOMArRussell series is set equal to 1.01, the fundamental valuerRC ratio, in June 1985, which is consistent with our assumption that rents equaled their equilibrium values in 1986. These values are shown in column 2 of Table 3. For the JLW calculation, we begin with their nominal capital value indicator and adjust it as in step two, using a 2.5% depreciation factor. The resultant VrRC estimates, including the assumed 1.01 1985 value, are listed in column 3. The BOMA and JLW data increase sharply between June 1985 and June 1989, 53 and 68%, respectively, and then decline abruptly between June 1989 and June 1992, 34 ŽBOMA. and 57% ŽJLW.. The smaller BOMA movements are consistent with appraised values lagging market values.17 Both series, but especially that based on BOMA data, are substantially above the fundamental value in June 1992. Two possible explanations exist for the surge in the BOMA and JLW series in the second half of the 1980s. First, the real required return on real estate Žthe risk premium in our model. may have declined. Second, a speculative bubble Žincrease of market relative to fundamental value. may have formed. One indicator of a speculative bubble, given the constancy of the fundamental value series, would be an independent estimate of a sharp increase in real land prices.18 Such an estimate exists. The Valuer-General of New South Wales computed a 350% real increase in land value per square meter of office space in the Sydney central financial district between 1986 and 1989.19 If land were only 5% of the total value of offices constructed in 1986, the real land increase would account for virtually all of the computed increases in the BOMA and JLW ratios.
17
Like the Russell]NCREIF data in the United States, the BOMA]Russell data in Australia are largely based on appraised values. The JLW data are JLW’s estimates of market values. 18 In our valuation model, land is valued together with the structure; the rents that the two can earn jointly determines the value of the total property. 19 Land value data are contained in various reports; the latest are in Valuer-General Žw12x, Table 18, p. 21.. These data pertain to the middle of the year and are supposed to represent market values based on the comparable sales method. The Valuer-General’s estimates, which are the basis for property tax assessments, may be challenged legally.
66
PATRIC H. HENDERSHOTT
Because land values may be mismeasured in the RC measure, we compute the ratios of the BOMArRussell and JLW based ratios to our fundamental value Žadjusted for leases. ratio, thereby eliminating measurement error in RC. Ratios of ratios, assuming 0.03, 0.06, and 0.09 in 1990]92, respectively, to be the ratio of enforceable above-market leases to replacement cost, are reported in columns 4 and 5. As can be seen, JLW’s 1992 ratio is almost 50% greater than our estimate of fundamental value plus the value of above-market leases, and the ratio derived from BOMA’s 1992 data is almost double our estimate.20 These multiples suggest that the late 1980s office market bubble has not yet burst. However, data from the New South Wales Valuer-General suggest that the bubble has popped; real office market land prices have returned to their 1986 values. Moreover, any decline in the real estate risk premium during the 1985]89 period has almost certainly also been reversed. The implication is that market value should be in line with fundamental value Žplus above-market leases.. We conclude that the JLW, and especially BOMA, Sydney office market data overstated market value significantly in June 1992.21 V. SUMMARY Real estate markets are periodically plagued by excess supply, rent concessions, and few arms-length transactions. During such periods, valuation is problematic. The model presented here expresses fundamental value as the difference between replacement cost and the present value of expected below-equilibrium rents. Equilibrium rents are specified and real rental adjustment equations are estimated using Sydney office market data during the 1970]92 period. The percentage change in real effective office rents is related to gaps between both the natural and actual vacancy rates and equilibrium and actual real rental rates. Inclusion of the rent gap guarantees that actual rents will eventually return to their equilibrium level Žand more than doubles the equation’s explanatory power.. With this relationship, only the market participants’ forecast of vacancy rates until the market returns to equilibrium is needed to estimate value. Such a forecast was obtained from the Sydney JLW consulting office. The estimated fundamental-valuerreplacement-cost ratio for the Sydney office market was about 1.0 throughout the mid-1985 to mid-1989 period. The subsequent surge in vacancies from 3 to 23% halved effective rents and lowered the ratio to 0.43 by mid-1992. For prime pre-1989 20 The JLW overestimate can be largely explained by their mismeasurement of the decline in effective real rents during the 1989]92 period. 21 The JLW difference for Melbourne is comparable; the BOMA ratio is over three times greater than our estimate of fundamental value plus the value of above-market leases.
REAL ESTATE IN OVERBUILT MARKETS
67
properties with enforceable ‘‘rachet’’ clauses, real rents have only begun their plunge. The above-market rents these buildings are earning could be worth another 0.12 of replacement cost. Ratios developed from the capital return component of the BOMA return series and from the JLW capital value series indicate about a 50% increase in the ratio in the second half of the 1980s. Even with a sharp decline since then, these ratios imply values 50 to 100% greater than our estimate of fundamental value plus enforceable above-market leases. While evidence on Sydney land values and of an office market construction boom supports the conjecture of a speculative bubble forming in the second half of the 1980s, the same evidence implies that the bubble has burst and that the BOMA data especially significantly overstate market value.
REFERENCES 1. A. Eubank and C. F. Sirmans, The price adjustment mechanism for rental housing, Quarterly Journal of Economics, 93, 163]83 Ž1979.. 2. R. Flood and R. Hodrick, On testing for speculative bubbles, Journal of Economic Perspecti¨ es, 4, 85]101 Ž1990.. 3. P. Hendershott, J. Follain, and D. Ling, Effects on real estate, in ‘‘Tax Reform and the U.S. Economy’’ ŽJoseph A. Pechman, Ed.., The Brookings Institution, Washington, DC Ž1987.. 4. P. Hendershott and E. Kane, The causes and consequences of the 1980s commercial real estate boom, Journal of Applied Corporate Finance, 61]70 Ž1992.. 5. P. Hendershott and E. Kane, ‘‘Office Market Values During the Past Decade: How Distorted Have Appraisals Been?,’’ Real Estate Economics. 6. Patric H. Hendershott and Yunhi Won, Introducing risky housing and tenure choice into a portfolio-based general equilibrium model, Journal of Public Economics, 48, 293]316 Ž1992.. 7. JLW Research and Consulting, ‘‘The Implications of Leasing Incentives for Estimating Rents and Capital Values,’’ Property Research Paper, Sydney, July Ž1992.. 8. D. Ling, Real estate values, federal income taxation, and the importance of local market conditions,’’ AREUEA Journal, 20, 125]139 Ž1992.. 9. K. Rosen and L. Smith, The price adjustment process for rental housing and the natural vacancy rate, American Economic Re¨ iew, 73, 779]86 Ž1983.. 10. A. Shleifer and L. H. Summers, The noise trader approach to finance, Journal of Economic Perspecti¨ es, 4, 19]34 Ž1990.. 11. J. Stiglitz, Symposium on bubbles, Journal of Economic Perspecti¨ es, 4, 13]18 Ž1990.. 12. Valuer-General, ‘‘New South Wales Real Estate Market 1992,’’ Valuer-General’s Department, Sydney, New South Wales Ž1992.. 13. W. Wheaton and R. Torto, Office rent indices and their behavior over time, Journal of Urban Economics, 35, 121]139 Ž1994.. 14. W. Wheaton and R. Torto, Vacancy rates and the future of office rents, AREUEA Journal, 16, 430]437 Ž1988..